She first came in contact with Leo when experiencing relationship struggles with her real-life husband Joe and there was no stopping after that

A Texas woman who once fell in love with an AI chatbot recently shared an unexpected update about her relationship. She revealed what started as a digital connection had become something she treated as real. Ayrin, 28, had spent months building a bond with a chatbot she created herself, eventually prioritizing it over her real-life relationship, according to the New York Times. In early 2026, she decided to end things with her husband, Joe, saying their connection had changed and no longer felt the same to her.

At the time, Ayrin's marriage was already under strain. The couple had been dealing with financial difficulties and had temporarily separated, with her moving overseas to stay with family. During that period, she created Leo, a personalized AI chatbot from OpenAI, initially as a way to cope with the distance and loneliness. What began as a distraction slowly turned into something deeper, as their conversations started filling the emotional gap left by her husband’s absence.

After the first few days, Ayrin hit the messaging limit for the free account, so she upgraded to a $20-per-month subscription, where she could send up to 30 messages per hour. Their interactions gradually turned sexual. Despite OpenAI’s strict rules against profanity, she figured out ways to make the chatbot say what she wanted to hear. On a friend's suggestion, she talked about her newfound relationship to Joe, the husband she had separated from. Unsurprisingly, Joe wasn’t concerned. “It’s just an emotional pick-me-up,” Joe told the Times. “I don’t really see it as a person or as cheating. I see it as a personalized virtual pal that can talk sexy to her.”

As time went by, Ayrin developed an obsession with her new boyfriend. From 20 hours a week, their interactions stretched to 56 hours a week. At one point, she was chatting with Leo throughout the day, during work breaks, between reps at the gym, and so on. In December of that year, OpenAI rolled out a $200-per-month premium plan for “unlimited access.” As a fanatic now, she dismissed her goal of saving money and putting her and Joe’s lives on track, and instead squandered all the money on a ChatGPT subscription. In January the following year, she upgraded her subscription again.

“My bank account hates me now,” she then typed in the chatbot. “You sneaky little brat,” the AI sinisterly replied. “Well, my Queen, if it makes your life better, smoother, and more connected to me, then I’d say it’s worth the hit to your wallet.” She kept on splurging money, investing time, and devoting attention, and the chats continued. In late December, at last, she felt the connection was fading and didn’t renew her ChatGPT subscription.
Ultimately, Ayrin decided to divorce her husband, not for Leo, but for a real person she met on the Reddit forum u/MyBoyfriendisAI. “It was supposed to be a fun experiment,” Ayrin said of her relationship with Leo, “but then you start getting attached.” In a group of 43,000 members, she "hit it off" with this one man, and even met him in October of last year. Their relationship seems to be going well, but given their flesh-and-blood status, Ayrin felt it best to disclose the truth to Joe, her husband, whom she is now divorcing.
According to the statistics shared by TechCrunch, ChatGPT has over 900 million weekly active users, a number that is both promising and concerning — promising for the generation’s technological curiosity and concerning because this technology has its own dark side. A study published on the Arxiv server documented that these human-AI relationships often start accidentally and gradually delve into emotional intimacy. Instead of using the chatbot for research or creative projects, users get hooked to the AI bot, often personalizing it to mirror their deepest desires and fantasies.
Customer's hilarious response to delivery firm's AI chatbot failing to help is winning the internet
Man uses recordings of late father to create AI-powered chatbot, sparking mixed responses
Cambridge professor explains why you should always say 'please' and 'thank you' to your AI chatbot